Rush orders wrecking your factory schedule? First decide who can insert them and who pays

"We finish the schedule, one phone call inserts a rush order, and the whole week starts over." The chaos from rush orders looks like a scheduling problem. It is actually a problem of authority and cost that nobody wrote down. This post breaks down the three prerequisites a factory scheduling system really needs: accurate standard times, a real definition of capacity, and knowing where work-in-progress is right now. It covers what happens if you implement a system before those are in place, and gives you a set of rush-order rules that cost nothing and that you can start writing this week.

Johnny Yang12 min read
Rush orders wrecking your factory schedule? First decide who can insert them and who pays

"We lay out the schedule on Monday. On Tuesday one phone call inserts a rush order, and the whole week starts over."

The person who said this in the meeting room that day was a plant manager. When he finished, he turned his laptop around to show me. On the screen was a very wide Excel sheet. Columns were dates, rows were machines, and the cells held work order numbers, carefully color-coded. He said this sheet takes two hours to build every Monday morning, and by Wednesday nobody is usually looking at it anymore. The owner sitting next to him added: "So we want to bring in a scheduling system and let it do the calculating."

I'm Johnny Yang, Brightstream Technology, Taiwan. I've led implementation teams for seven years, handled 252+ projects across 32+ industries. I've had this conversation many times, so I'll say it plainly up front: Rush orders are not a problem a scheduling algorithm can solve. What a scheduling system does is "calculate a feasible sequence under given conditions." It cannot, at any point, answer "should this order be inserted at all"—that is a business judgment, not a math problem. When a factory gets thrown into chaos by rush orders, the real cause is usually two things. First, nobody has written down who has the authority to decide on a rush order, or who bears the delay it causes. Second, you don't have data that can calculate the cost. The standard times were estimated three years ago, capacity is derived from "number of machines times shift hours," and nobody can say for sure which station the work-in-progress is stuck at right now. If any one of these three fails, the schedule the system produces is a professional-looking guess. Buying a system will not turn a guess into a fact. It only makes the guess run faster and harder to question.

First, you're right: wanting to insert rush orders is not wrong

I know that at this point many owners feel a bit uncomfortable, thinking I'm saying rush orders are bad. They're not.

Companies that insert rush orders usually do it because they can win urgent business. A customer adds on at the last minute, a major client asks to move up a date, or one order's margin is clearly better than the ones scheduled ahead of it—in those cases inserting the order is the right business decision, and not doing it would be stupid. In the manufacturing projects I've seen, more than half of revenue growth often comes from this kind of flexibility. You're more willing to take the job than your competitors, more willing to say yes. That is your competitive edge, not your disease.

The problem is that the act of inserting an order has been cut off from its consequences. The person who decides is in the office. The people who bear the consequences are on the floor. A salesperson promises the customer delivery three days early, hangs up, walks over to production control and mentions it. The production planner's face drops, but he schedules it anyway. Then the three orders that were in that slot get pushed back. One of them belongs to another salesperson's customer, and that customer will call next week to yell at someone. Along this entire chain of cause and effect, not a single step gets recorded.

So at the end-of-month review, what you hear are abstract lines like "delivery dates have been tight lately" and "the floor isn't cooperating enough." Nobody can point to a sheet and say: this month we inserted rush orders this many times, here's who decided each one, here's how many orders got delayed in total, and here's what happened to those delayed orders. These numbers don't exist. Not because nobody calculated them, but because the raw data was never kept in the first place.

A scheduling system has three real prerequisites

I'll lay these three out clearly so you can check them against your own factory directly. If you can't answer these three questions, don't buy a scheduling system yet.

Question one: Where do your standard times come from, and when were they last updated? The core input of a scheduling system is "how long does this operation take." If that number was estimated from a veteran operator's memory, or filled in during quotation to calculate cost, then every cell the system schedules is built on that number. I often run a very simple test on site: pick three part numbers at random, ask production control what the standard time is, then ask the person who actually does that operation. Where the gap is over 20%, it's usually a lot more than 20%. Worse, changeover time and waiting time are often not recorded at all. So the system thinks a machine can take the next order the moment it stops, when in reality die changes, machine adjustment and first-article inspection alone eat most of a shift.

Question two: How do you define capacity? Many factories derive capacity from the number of machines times shift hours. This looks beautiful on paper and doesn't survive a single day on the floor. Real capacity gets eaten by three things: people (the operator for that machine is on leave today, so the machine is there but the capacity isn't), dies and fixtures (two orders fighting over the same die, so the machine sits idle and still can't run), and inspection and rework (finished doesn't mean shippable—does the time stuck in inspection count?). If your definition of capacity doesn't account for these three, the scheduling system will happily produce a fully loaded schedule for you, and the floor will miss it every single day.

Question three: Can you say right now which station your work-in-progress is at? This is the most practical of the three. For a scheduling system to reschedule, it has to know the current state—which orders have finished which operations, which are stuck at station three, which are waiting for subcontractors to send parts back. If getting this information means someone walking the floor and asking, then the system's rescheduled result will always lag reality by half a day to a full day. And floor-level production reporting is a pit of its own: in many factories the numbers on the reporting sheet are filled in all at once before going home, and the sequence and times don't match what actually happened. I broke down how to diagnose this in "Production reporting gets done, but the numbers still can't be trusted?", so here's just the conclusion—if your reporting data isn't real, your scheduling system is blind in one eye.

PrerequisiteWhat happens without itQuick self-check
Accurate standard times, including changeoverThe system produces a nice-looking schedule, the floor falls behind daily, and eventually everyone goes back to the Excel sheetPick three part numbers at random and ask production control and the floor operator how long the same operation takes
Capacity net of people, dies and inspectionThe system loads machines to 100%, actual utilization never catches up, and supervisors start "manually fine-tuning"Take any day last month and ask how many machines sat idle that day because of missing staff or die conflicts
Real-time WIP locationRescheduling is based on yesterday's state, and the impact of a rush order can't be calculatedCall right now and ask: what station is work order No. 3 at, and how long does it take to get an answer

What happens if you implement before these three are in place

Among the failed implementations I've seen, factory scheduling systems die in a particularly consistent way, because they die fast and visibly.

Stage one is distrust. The floor takes one look at the first schedule the system produces and knows it's impossible. Maybe it puts two orders that need the same die in the same time slot, or it counts changeover time as zero. At this point the plant manager does something completely natural: he adjusts it by hand. After the adjustment, that schedule is no longer the system's schedule. It's the plant manager's schedule that happens to live on the system's screen.

Stage two is running two tracks. Because the system's schedule can't be used directly, production control starts maintaining two things—one in the system for the owner to look at, and an Excel file on their own computer for the floor. At this stage you'll see a classic scene on site: the sales support staff still has that Excel file open on screen, and it's more current than the one in the system. By now, the money you spent has become a pure report generator.

Stage three is blame. The owner finds the system didn't reduce rush orders, and the conclusion usually lands on "this system is no good" or "the consultants weren't professional." Sometimes that's true. More often, the three prerequisites were never filled in from start to finish, and during implementation nobody was willing to say so plainly—because saying it plainly means the project gets delayed, and nobody likes delays.

I want to point out something that's easy to misunderstand. In the process above, the most expensive cost isn't the money spent on the system. It's the floor losing trust in the word "system." With the same group of people, any time you try to push anything digital three years later, the first thing you'll hit is "that's what they said last time too." Earning that trust back is much harder than the first implementation.

What you write down first costs nothing: who can insert orders, and who pays

Before you decide whether to buy a system, there's something you can do first, and you can do it on paper.

The first thing is to grade rush orders. Not all rush orders are the same. One kind is a customer adding more of the same product—same upstream and downstream processes, just higher volume. Its impact on the schedule is small. Another kind is a completely different part number being inserted, requiring a die change, machine adjustment, maybe a last-minute material pull. That eats more than just machine time. The third kind is the most troublesome: inserting the order while also requiring that none of the original orders slip. Strictly speaking, that isn't inserting an order, it's overtime—and who decides to spend money on overtime has to be made clear first. Once you sort rush orders into these types, the act of sorting alone turns the discussion that follows from emotion into fact.

The second thing is to write down decision authority. Which level of person can decide which type of rush order—this needs to be on paper, with specific thresholds. The thresholds can be rough at first. What matters is that they exist. In most factories I've seen, the situation is "the salesperson tells production control and that's it," and production control actually has no power to refuse, because if they refuse and the customer walks, the blame lands on them. If you don't fix this question of accountability, ten different systems will all end the same way.

The third thing, and the one the fewest people do: record the consequences of every rush order. Nothing complicated, one sheet of paper. Columns: date, which order was inserted, who decided, reason, which orders were pushed back, and by how many days. Record it for two straight months and you'll have something more convincing than any system report. In the projects I've seen, the same result usually shows up—rush orders come heavily from a few customers or a few salespeople, and the orders that get sacrificed also often belong to the same group of customers. The value of this paper record is that it turns the vague complaint "everybody's inserting orders" into a concrete fact you can take into a meeting and take to a customer.

After doing these three things, if you still need a scheduling system, you'll know what you're buying and what questions to ask vendors. The more common outcome is this: once you've done them, you discover that the thing you should fix first isn't scheduling at all, but somewhere else. For how to decide which process to fix first, I have a set of criteria in "Which process should you digitize first?" you can refer to.

So when should you actually buy a scheduling system?

There are times you should buy one. I don't want this to read like scheduling systems are never necessary.

If your standard times are maintained, your capacity definition accounts for people and dies, and floor reporting is real-time and truthful, then you'll hit a different kind of pain: too many constraints for a human brain to schedule. A dozen-plus machines, seven or eight operations, a different routing for every order, some operations that only specific machines can do, plus delivery dates and changeover costs to weigh together—that kind of combinatorial explosion really is beyond what Excel can handle. At that point the value of a scheduling system is real.

There's one more situation where you should buy: your scheduling knowledge currently exists only in one person's head, and that person is about to retire. In that case the point of buying a system isn't to calculate better than he does. It's to force your team to turn the rules in his head into parameters that can be checked. That process itself is worth more than the system.

On the other hand, if your pain right now is "nobody follows the schedule once it's made," that's usually not a scheduling problem. It's an execution problem. The floor usually has very specific reasons for not following it. Those are worth asking about first, instead of handing them a smarter schedule to not follow.

So what should you do tomorrow

Do just one thing tomorrow morning: take a sheet of A4 paper and tape it to the wall next to production control. Write five columns—date, which order was inserted, who decided, which orders were pushed back, how many days. Make it clear to production control that this isn't about catching anyone out. It's just to write things down first, and in two months we'll look at it together.

This takes no meeting, no budget, and no waiting for anyone's approval. In two months you'll have something that lets you answer three questions you can't answer now: how often rush orders actually happen, who the cost falls on, and whether the orders those costs bought were worth it. If the answer is yes, what you need to deal with is the mechanism for capacity and delivery commitments. If the answer is no, what you need to deal with is decision authority. Neither path starts with buying a system.

If after two months you can't make sense of what that sheet is telling you, or you're not sure whether your standard times can actually be used as system inputs, you can talk to me for thirty minutes, and I'll go through those three prerequisites with you one by one.

Found this useful? Pass it on
30 minutes · free

Bring the one thing you are stuck on — a free 30-minute diagnostic

You do not need to know what system you want. Tell Johnny Yang about the one thing that keeps blocking you, and in 30 minutes you will know whether to skip it, build it with coaching, or pay for a custom build. No pitch — if we cannot help, you will hear that straight.

Johnny Yang replies himself · limited spots each month · or just book a free 30-minute diagnostic

Johnny Yang (熊董)
Johnny Yang (熊董)
Founder, Brightstream Technology

Serial founder leading a small team that builds systems into an owner's everyday work.

As a founder
7 yrs
Projects delivered
252+
Industries served
32+
About Johnny